task_categories:
- translation
language:
- ar
- en
size_categories:
- 10K<n<100K
license: cc-by-4.0
Introduction
This dataset represents a comprehensive collection of parallel Arabic-English texts from the Thaqalayn Hadith Library, a premier source for exploring the classical hadith tradition of the Imāmī Shia Muslim school of thought. The library focuses on making primary historical sources accessible, serving as a bridge between past wisdom and contemporary study. The dataset features translations of significant classical Imāmī hadith texts, allowing for a deep dive into the linguistic and cultural heritage of this era.
Content Details
The Thaqalayn Hadith Library includes Arabic-English parallel texts from the following classical collections:
- Al-Kāfi (The Sufficient)
- Muʿjam al-Aḥādīth al-Muʿtabara (A Comprehensive Compilation of Reliable Narrations)
- Al-Khiṣāl (The Book of Characteristics)
- ʿUyūn akhbār al-Riḍā (The Source of Traditions on Imam al-Riḍā)
- Al-Amālī (The Dictations) by Shaykh Muḥammad b. Muḥammad al-Mufīd
- Al-Amālī (The Dictations) by Shaykh Muḥammad b. ʿAlī al-Ṣaduq
- Al-Tawḥīd (The Book of Divine Unity)
- Kitāb al-Ḍuʿafāʾ (The Weakened Ones)
- Kitāb al-Ghayba (The Book of Occultation) by Abū ʿAbd Allah Muḥammad b. Ibrāhīm al-Nuʿmānī
- Kitāb al-Ghayba (The Book of Occultation) by Shaykh Muḥammad b. al-Ḥasan al-Ṭūsī
- Thawāb al-Aʿmāl wa ʿiqāb al-Aʿmāl (The Rewards & Punishments of Deeds)
- Kāmil al-Ziyārāt (The Complete Pilgrimage Guide)
- Faḍaʾil al-Shīʿa (Virtues of the Shīʿa)
- Ṣifāt al-Shīʿa (Attributes of the Shīʿa)
- Maʿānī al-ʾAkhbār (The Meanings of Reports)
- Kitāb al-Muʾmin (The Book of the Believer)
- Kitāb al-Zuhd (The Book of Asceticism)
- Nahj al-Balāgha (The Peak of Eloquence)
Purpose and Application
The dataset aims to showcase the unmatched literary quality of Classical Arabic, distinguished from Modern Standard Arabic, particularly in its preservation from the European translation trends of the 19th and 20th centuries:
- Refinement of Machine Translation (MT): The complex grammatical structures and rich lexicon of Classical Arabic present a unique challenge for MT systems, pushing the boundaries of translation accuracy and fluency.
- Development of Language Models: These texts serve as a foundation for training sophisticated Large Language Models (LLMs) capable of understanding and replicating the depth of Classical Arabic.
- Preservation of Linguistic Heritage: This dataset preserves the original form of Classical Arabic, providing a standard of excellence against which modern writings can be compared.
Suggested Research Application: Iterative Translation Refinement
A notable application of this dataset is the enhancement of contemporary Arabic writing through back-translation. Existing models can back-translate English texts into Arabic, potentially producing a less sophisticated form. This offers an opportunity to:
- Generate Imperfect Arabic Corpus: Use back-translation to create a corpus of Arabic text that is less refined than the original Classical Arabic.
- Train Refinement Models: Develop models that refine the imperfect Arabic by comparing it to the original texts, aiming to restore the classical eloquence.
- Enhance Contemporary Arabic Writing: Apply these models to modern Arabic texts, elevating their literary quality by infusing classical stylistic elements, making the language resonate with its classical roots.
Credits
Credits go the Thaqalayn website for their compilation of Arabic and English texts. Also, the original webscrape is done by jenusi on GitHub in this repo. I only compiled it in the form of two columns for texts from all books. I also converted the numbers from western Arabic (0123456789) to eastern Arabic (٠١٢٣٤٥٦٧٨٩).